The present invention relates generally to storage systems and, more particularly, to distributed storage system and method for managing data recovery.
In a distributed storage system, methods for replicating data to other nodes and failure recovery of a storage node are known. The distributed storage system performs to mirror twice or triple replication data to recover composited disks, nodes and/or network failures.
When many nodes fail, the distributed storage system changes the layout of replication data which are stored to the failure nodes to healthy nodes and it copies data to healthy nodes to keep redundancy. However, if too many nodes have failed, physical capacity is insufficient and replication or host write will fail.